[NNTP] [2501] Reader commands in transit servers and vice-versa

[Russ Allbery]

Clive D W Feather <clive at demon.net> writes:

> Issue: if a server advertises itself as transit-only, should it be
> allowed to implement some of the reader-only commands or not?

Yes.  As Andrew pointed out, there's no reason to forbid a server from
doing this, and it's useful for various purposes.

The question is more whether we want to define a mechanism for the server
to *tell* a client that it's doing this.  If we don't, then a compliant
client will never know it can use the commands, won't use them, and
therefore they'll be there basically only for debugging and for internal
purposes.

Personally, I think that's fine, and better than the complexity of trying
to define a way of communicating all this information.  I would rather
take the route of specifying what the *client* should do based on what the
server says it supports and simply remain mum on the question of what the
server might do outside of what it's advertising that it does.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>